The compliance foundation
HIPAA with BAAs across the stack
Business Associate Agreements are in place across the cloud infrastructure—including the AI model provider.
PHI stays in the covered environment
AI models run through AWS Bedrock inside the covered environment. Clinical data never leaves for a public AI service.
Encrypted everywhere
All data is encrypted in transit and at rest with AES-256 encryption.
Isolated by tenant
Row-level security enforces strict practice-level data isolation. Every access is authenticated.
No public database access
The production database is not publicly reachable; administrative access requires an audited bastion session.
Auditable by construction
Scores and configurations are append-only and versioned—any past decision can be reconstructed exactly as it was made.
How the clinical AI is governed
Trust is earned structurally—by being groundable, reproducible, and governed—rather than asked for.
Minimum necessary output
The AI extracts only codes from a controlled clinical vocabulary plus the specific verbatim sentence supporting each finding—never free-form narrative about the patient.
No fabrication by design
Uncertainty is representable: vague durations keep the clinician’s original words, and severity is never inferred from questionnaire scores—the system cannot silently invent clinical facts.
Payer-blind scoring
Insurance coverage is excluded from candidacy scoring and from every display panel by ratified principle—a patient’s clinical priority is never shaped by their coverage.
Human governance
No model version or scoring weight reaches patient-facing workflows without clinician ratification—new versions run in shadow beside the live system until clinicians approve them against real cases.
Questions about security or compliance?
We’re happy to walk your security or compliance team through the architecture.